Re: [bug] Adding -z to git status seems to disable relative path

From: pp yy <yoann.mac.donald@gmail.com>
Date: Dec 18, 2019, 14:04 UTC
Message-ID: <CAK60LiV_e=q33R20WJU=Hxrn6kKuXwehM+9j_8cmWX4krnmv2w@mail.gmail.com>
In-Reply-To: <20191216063425.GA2127604@coredump.intra.peff.net>

Hi Peff

Even without adding -s the option "status.relativePaths" isn't respected. $ git -c status.relativePaths=true status -z test ?? plugins/git/test

It is explained in the man in fact.
Porcelain v1 is applied when using"-z" and porcelain v1 explicitly state:
"
1. The user’s color.status configuration is not respected; color will
always be off.
2. The user’s status.relativePaths configuration is not respected;
paths shown will always be relative to the repository root.
"

With this in mind, it seems impossible to have the same output as `git
-c status.relativePaths=true -c color.status=always status -s test`
but with records separated by '\0' (for scripting purpose), and that's
what I was looking for :
- list files status
- separated by '\0'
- with colors

Regards,

Le lun. 16 déc. 2019 à 07:34, Jeff King <peff@peff.net> a écrit :

Show 69 quoted lines

>
> On Sun, Dec 15, 2019 at 09:31:23PM +0100, pp yy wrote:
>
> > Maybe i missed something but i can't get relativepath when adding '-z'
> >
> > $ git --version
> > git version 2.17.1
> > $ git status -s test
> > ?? test
> > $ git status -s -z test
> > ?? plugins/git/test
> > $ git -c status.relativePaths=true status -s test
> > ?? test
> > $ git -c status.relativePaths=true status -s -z test
> > ?? plugins/git/test
> >
> > Bug or did i missed something ?
>
> I think it's a bug. At first I thought you were running up against the
> implied porcelain output:
>
>   -z
>       Terminate entries with NUL, instead of LF. This implies the
>       --porcelain=v1 output format if no other format is given.
>
> but you are explicitly saying "-s" (and running this in a terminal shows
> that the result is colorized, which means we're triggering the short
> format and not porcelain).
>
> And indeed, the "-z" code path seems to ignore the prefix entirely.
> Something like this would fix it:
>
> diff --git a/wt-status.c b/wt-status.c
> index cc6f94504d..3a0e479407 100644
> --- a/wt-status.c
> +++ b/wt-status.c
> @@ -1837,9 +1837,13 @@ static void wt_shortstatus_status(struct string_list_item *it,
>                 putchar(' ');
>         putchar(' ');
>         if (s->null_termination) {
> -               fprintf(stdout, "%s%c", it->string, 0);
> +               struct strbuf scratch = STRBUF_INIT;
> +               fprintf(stdout, "%s%c",
> +                       relative_path(it->string, s->prefix, &scratch), 0);
>                 if (d->rename_source)
> -                       fprintf(stdout, "%s%c", d->rename_source, 0);
> +                       fprintf(stdout, "%s%c",
> +                               relative_path(d->rename_source, s->prefix, &scratch), 0);
> +               strbuf_release(&scratch);
>         } else {
>                 struct strbuf onebuf = STRBUF_INIT;
>                 const char *one;
>
> Now I do think it's a little weird to use "-z" with "--short" in the
> first place, since the whole point of "-z" is make something that can be
> parsed. And the whole point of "--short" versus "--porcelain" is that
> the latter is stable and predictable (so it doesn't respect config at
> all).
>
> But I guess I could imagine a use case where a script is taking in paths
> from the "-z" and then showing them to the user without further
> processing (and so it's convenient to get the relative path directly
> rather than computing them itself). I do wonder if there are any other
> surprises in "--short" that might be lurking, though (IIRC, we do not
> even promise that the output will stay syntactically the same between
> versions; it could change to something completely different, or add new
> lines in future versions).
>
> -Peff